arXiv subjects

Julien Mozziconacci

Publications and source records attributed to Julien Mozziconacci.

9 recordsLinked to original sources

arXiv2026

DNA Language Models: An Assessment of Pre-Training for Fine-Tuning Tasks

Recent breakthroughs in foundation models and Large Language Models (LLMs) have introduced new opportunities for studying and decoding genomic sequences. Several state-of-the-art approaches, such as DNABERT2, rely on transformer-based architectures, while others, such as ConvNova, still build upon more conventional convolutional models. However, systematic benchmark comparisons across these methods remain scarce. Given that transformer-based models require extensive and costly pretraining, it is crucial to evaluate whether their performance gains justify this overhead. Moreover, LLMs such as DNABERT2 typically rely on Byte Pair Encoding (BPE) tokenization, whose relevance for DNA sequence representation is still debated within the genomics community. In this work, we investigate three key questions: (i) do transformer-based models provide sufficient improvements on fine-tuning tasks upon heavy pretraining, (ii) what is the actual contribution of pretraining in this setting, and (iii) how does BPE tokenization impact performance on genomics-related tasks?

q-bio.GN↗

arXiv2019

The 3D genome shapes the regulatory code of developmental genes

We revisit the notion of gene regulatory code in embryonic development in the light of recent findings about genome spatial organisation. By analogy with the genetic code, we posit that the concept of code can only be used if the corresponding adaptor can clearly be identified. An adaptor is here defined as an intermediary physical entity mediating the correspondence between codewords and objects in a gratuitous and evolvable way. In the context of the gene regulatory code, the encoded objects are the gene expression levels, while the concentrations of specific transcription factors in the cell nucleus provide the codewords. The notion of code is meaningful in the absence of direct physicochemical relationships between the objects and the codewords, when the mediation by an adaptor is required. We propose that a plausible adaptor for this code is the gene domain, that is, the genome segment delimited by topological insulators and comprising the gene and its enhancer regulatory sequences. We review recent evidences, based on genome-wide chromosome conformation capture experiments, showing that preferential contact domains found in metazoan genomes are the physical traces of gene domains. Accordingly, genome 3D folding plays a direct role in shaping the developmental gene regulatory code.

q-bio.GN↗

arXiv2019

Turing-like patterns in an asymmetric dynamic Ising model

To investigate novel aspects of pattern formation in spin systems, we use a mapping between reactive concentrations in a reaction-diffusion system and spin orientations in a dynamic multiple-spin Ising model. While pattern formation in Ising models always rely on infinite-range interactions, this mapping allows us to design an finite-range interactions Ising model that can produce patterns observed in reaction diffusion systems including Turing patterns with a tunable typical length scale. This model has asymmetric interactions and several spin types coexisting at a site. While we use the example of genetic regulation during embryo-genesis to build our model, it can be used to study the behavior of other complex systems of interacting agents.

nlin.AO↗

arXiv2018

Kinetic signature of cooperativity in the irreversible collapse of a polymer

We investigate the kinetics of a polymer collapse due to the formation of irreversible crosslinks between its monomers. Using the contact probability $P(s)$ as a scale-dependent order parameter depending on the chemical distance $s$, our simulations show the emergence of a cooperative pearling instability. Namely, the polymer undergoes a sharp conformational transition to a set of absorbing states characterized by a length scale $\xi$ corresponding to the mean pearl size. This length and the transition time depend on the polymer equilibrium dynamics and the crosslinking rate. We confirm experimentally this transition using a DNA conformation capture experiment in yeast.

cond-mat.soft↗

arXiv2016

The physics of epigenetics

In higher organisms, all cells share the same genome, but every cell expresses only a limited and specific set of genes that defines the cell type. During cell division, not only the genome, but also the cell type is inherited by the daughter cells. This intriguing phenomenon is achieved by a variety of processes that have been collectively termed epigenetics: the stable and inheritable changes in gene expression patterns. This article reviews the extremely rich and exquisitely multi-scale physical mechanisms that govern the biological processes behind the initiation, spreading and inheritance of epigenetic states. These include not only the changes in the molecular properties associated with the chemical modifications of DNA and histone proteins, such as methylation and acetylation, but also less conventional ones, such as the physics that governs the three-dimensional organization of the genome in cell nuclei. Strikingly, to achieve stability and heritability of epigenetic states, cells take advantage of many different physical principles, such as the universal behavior of polymers and copolymers, the general features of non-equilibrium dynamical systems, and the electrostatic and mechanical properties related to chemical modifications of DNA and histones. By putting the complex biological literature under this new light, the emerging picture is that a limited set of general physical rules play a key role in initiating, shaping and transmitting this crucial "epigenetic landscape". This new perspective not only allows to rationalize the normal cellular functions, but also helps to understand the emergence of pathological states, in which the epigenetic landscape becomes dysfunctional.

physics.bio-ph↗

arXiv2007

Nucleosome Chiral Transition under Positive Torsional Stress in Single Chromatin Fibers

Using magnetic tweezers to investigate the mechanical response of single chromatin fibers, we show that fibers submitted to large positive torsion transiently trap positive turns, at a rate of one turn per nucleosome. A comparison with the response of fibers of tetrasomes (the (H3-H4)2 tetramer bound with ~50 bp of DNA) obtained by depletion of H2A-H2B dimers, suggests that the trapping reflects a nucleosome chiral transition to a metastable form built on the previously documented righthanded tetrasome. In view of its low energy, <8 kT, we propose this transition is physiologically relevant and serves to break the docking of the dimers on the tetramer which in the absence of other factors exerts a strong block against elongation of transcription by the main RNA polymerase.

q-bio.BM↗

arXiv2007

A Physical Model for the Condensation and Decondensation of Eukaryotic Chromosomes

During the eukaryotic cell cycle, chromatin undergoes several conformational changes, which are believed to play key roles in gene expression regulation during interphase, and in genome replication and division during mitosis. In this paper, we propose a scenario for chromatin structural reorganization during mitosis, which bridges all the different scales involved in chromatin architecture, from nucleosomes to chromatin loops. We build a model for chromatin, based on available data, taking into account both physical and topological constraints DNA has to deal with. Our results suggest that the mitotic chromosome condensation/decondensation process is induced by a structural change at the level of the nucleosome itself.

q-bio.SC↗

arXiv2007

Structural plasticity of single chromatin fibers revealed by torsional manipulation

Magnetic tweezers are used to study the mechanical response under torsion of single nucleosome arrays reconstituted on tandem repeats of 5S positioning sequences. Regular arrays are extremely resilient and can reversibly accommodate a large amount of supercoiling without much change in length. This behavior is quantitatively described by a molecular model of the chromatin 3-D architecture. In this model, we assume the existence of a dynamic equilibrium between three conformations of the nucleosome, which are determined by the crossing status of the entry/exit DNAs (positive, null or negative). Torsional strain, in displacing that equilibrium, extensively reorganizes the fiber architecture. The model explains a number of long-standing topological questions regarding DNA in chromatin, and may provide the ground to better understand the dynamic binding of most chromatin-associated proteins.

q-bio.BM↗

arXiv2004

How does the chromatin fiber deal with topological constraints?

In the nuclei of eukaryotic cells, DNA is packaged through several levels of compaction in an orderly retrievable way that enables the correct regulation of gene expression. The functional dynamics of this assembly involves the unwinding of the so-called 30 nm chromatin fiber and accordingly imposes strong topological constraints. We present a general method for computing both the twist and the writhe of any winding pattern. An explicit derivation is implemented for the chromatin fiber which provides the linking number of DNA in eukaryotic chromosomes. We show that there exists one and only one unwinding path which satisfies both topological and mechanical constraints that DNA has to deal with during condensation/decondensation processes.

q-bio.SC↗